Predict Maintenance Decisions For Multi-Component System Based On Kernel Density Estimation

The purpose of prediction and health management is to utilize equipment operation monitoring data,evaluate and predict the possible future status of equipment,the time of faults,etc.,as a basis for formulating maintenance plans,improve maintenance efficiency,save maintenance costs,increase system reliability,and achieve intelligent maintenance decision management of the system.Prediction and health management,as one of the main applications of industrial big data,are of great significance in improving the reliability and efficiency of modern industrial systems.This paper constructs an adaptive kernel density estimation residual life prediction model for multi-component systems,formulates a maintenance decision stopping criterion considering variance,and conducts group maintenance for multicomponent systems.Further research on group maintenance decisions considering imperfect maintenance effects is carried out and verified through experiments.The main research content includes:(1)For a multi-component system that can monitor degradation data in real time,the influence of kernel density estimation kernel function and kernel density estimation window width on the results of kernel density estimation is analyzed,and an adaptive kernel density estimation method is used to establish the residual life prediction model of components.(2)Develop maintenance decision stopping criteria based on the predicted results,establish a group maintenance decision model considering economic correlation,and provide optimization algorithms.Firstly,considering the degree of deviation between the residual life variance and its mean,this paper develops a maintenance decision stopping criterion that considers the variance of the residual life probability distribution;Secondly,establish a long-term average minimum cost rate model and use genetic algorithm to optimize the optimal maintenance time of components;Subsequently,the group maintenance strategy is determined and a penalty function model is used to optimize the optimal group maintenance time of the system to achieve cost savings;Finally,the feasibility and accuracy of the proposed method were verified through experiments.(3)A group maintenance strategy considering the effect of imperfect maintenance is proposed to address the fact that most maintenance activities involve imperfect maintenance,which means that the degradation level of components cannot be restored to a completely new state through maintenance activities,and can only be restored to a certain state between the completely new state and the pre maintenance state.Firstly,a residual life prediction model considering imperfect maintenance effect based on adaptive kernel density estimation is constructed;Then,establish a imperfect group maintenance decision model with the minimum long-term average cost rate and provide an optimization algorithm;Finally,combined with numerical examples,the universality and effectiveness of the proposed maintenance strategy are verified.